Sneak peak - T5250 1.5GHz @ 1.8GHz

I have a few things to figure out still with vista -- but I successfully took the processor to 1.8GHz by making it run at 800FSB. This is not software done. I will be posting a full step-by-step guide in the next 2-3 days on how to do this yourself. So if you have a 667FSB processor in the Socket P Santa Rosa motherboard, you have the chance for a free speed boost courtesy of a little copper wiring

 

That sounds kinda dangerous :O... that said, nice work

 

Yay! Looking forward to your guide. Pinmod ftw.

 

Thought it was a VID problem because Vista kept crashing...removed mod, rebooted, same problem -- but a reinstall proved it was ok. So tomorrow I will do the mod with a fresh vista and if all goes well. Guide will be produced.

 

well, what happens if you have a 800mhz bus already? can you take it to 1066? haha =)

 

yes you could if the cpu and rest of the components could handle it..you could go as high as the parts will allow and be stable

 

Huh maybe ill try when my system comes! if mines at 800mhz standard does that mean it has a better chance of hitting 1066 than yours which is at 667mhz?

 

MAYBE.... some chips are more inclined to overclock then others are.. take for example the older pentium 4 northwood's.. the 2.4c i had went up to 3.6ghz in a huge overclock.. the same series 3.2c chip would only hit 3.5ghz then i replaced that 3.2 with a 3.0c and hit 4.1ghz using water cooling.. so it all dpends no two chips will always be exactly the same.. and be CAREFUL... if you oc to much or cause issues your warrantee is out the door,,,

 

Yeh mite hav to try that tho! wt u think?

 

You might have to try that though. What do you think?***

 

wtf r u chattin bout m8?

 

The 800FSB CPUS CANNOT DO 1066FSB!

Sorry to break the news to you guys. There are no BSEL mods capable in the socket P platform to run the CPU at 1066FSB. There is 400/667/800 and thats it

 

dude... try some of my ES CPU stuff.. still have some T5450 Engineer sample, with unlock multipliers for you to overlock.
PM me for more info. I also have a T7100 ES version also.. can you overclock to 2.5ghz hhahahah

 

So which pins I should jumper?

 

nice work, can't wait for your guide, what is your Windows rating on your cpu now that you've got it to 1.8ghz?

 

Just to keep you guys informed. Have not had the time to take it apart and redo the pins.
@HT2: The pins you must connect with a thin copper thread are the following:

This would be the view on the socket ^ Connect Column 22 B and C together and put the CPU CAREFULLY back in. You can be sure of the position using the missing pins at the top left. It will be oriented the same if you have the laptop facing you as you normally would. Its the furthest right corner just like the picture. As long as you don't screw anything else up while opening the laptop and ensure the pin's are ONLY touching the specified places there is almost no way you can fry the processor. It will either work or it wont and you CPU will function at the normal 667FSB.

The problem that was seeing is that it keeps being throttled to the 6x multiplier even though Windows and a few other App's are reading it (including the BIOS) as 1.8GHz. CPU-Z reads it as 1200MHz (6x200).

Some people have stated this happened the same with the Yonah processors when bumped from 533-667FSB.

 

And the Pentium M 400MHz -> 533MHz before that.

Good stuff ps2cho.

 

Well it was not successful because the multiplier keeps getting throttled to 6x and there are no programs/ways to bump it up via software. I have done some research and it seems it could be a problem with the chipset which nobody has found the answer to. I sure can't figure it out. I don't have the knowledge.

So for right now all I have done is figure out it can be pinmodded to 800FSB -- we just need someone with more technical knowledge to figure out why the 6x multiplier is getting stuck. Someone successfully got a Yonah to 667FSB without the stuck 6x multi but it was on an ATI chipset and that is why the chipset could be the culprit.

 

any update?

 

why? is it really worth destroying your cpu for this? if you want faster, order a faster processor!

 

because its human nature to always want more than what you paid for...

 

Some people enjoy living on the edge.

 

destroying the CPU?

isn't the t5250 just a handicapped t7100 that runs at 667 MHz FSB instead of 800MHz? Given the same voltage, the t5250 at 800 MHz FSB and the T7100 should have roughly the same temperatures. How is that destruction of the CPU?

 

because mess up the wrong pins and you fry it. idk thats just my opinion. i used to be so into "over clocking" years ago, though it was so cool to to have a 3.8 ghz processor when it was only 3.2 out of the box. but in honesty, it got so much hotter and it wasn't noticeably faster, only in benchmarks. do you think that there is a reason that it is handicapped? its like graphics cards. they have the lower end models that are just the high end models that cant pass the tests because something is wrong with a pipeline. they disable the pipeline to make it work and sell it at a lower price/lower performance. people think they are so smart to unlock those broken pipelines because on paper its the same as the faster more expensive card, but it really isn't. it can cause stability problems, and longevity problems. i wouldn't do it, but if your so inclined to go for it.

 

Ok guys this how it is right now:

Something in the chipset/OS is causing the CPU to throttle down to the 6x permanently. I have had another use confirm it on another laptop (not dell) and it does the same thing.

Until somebody can figure out a fix to change the multiplier it won't work. We have tried all programs to change the multi and nothing works.

So for right now we are out of luck.

 

wah wah, why not just use a program to oc it in windows?

 

Because then the Bus speeds are sent over the specified frequencies...possibly increasing the chance that your HDD and all peripherals attached to the system bus will go out of whack and die. The PCI/PCI-E Bus' are not locked, so increasing the FSB via software increases the FSB, and the other BUS speeds too.

I want to keep my HDD and the rest of the components in tact, a CPU is easily replaceable. HDD, motherboard, Graphics card are less replaceable because it would be almost impossible to determine if the GPU or the motherboard went wrong without having another GPU to test...and its a laptop...who carries spare GPU's.... :/

Just to update...nobody has found a solution to the throttling yet.

 

Have you tried using RMClock to manually switch to a higher multiplier?

 

Someone figure out how to change the divider (FSB : DRAM) ratio.

 

Just a nice bump to see if anybody here has any other ideas.

 

Are you sure the multiplier is permanently stuck at 6x? (1000mhz) I saw the same thing with my t5250 is CPU-z, but whenever I utilized the processor a good bit the multiplier would kick up to 9x.

I'm not sure exactly how the power management feature works, since the multiplier drops in my barebones XP install as well as Vista that came with my laptop.

 

What if you set the power options in Windows to Maximum performance to keep it running at max speed? Have you tried?

 

You guys think that I wouldn't know what speedstep was when I am pulling out the processor and pinmodding

No its not speedstep lol. Its chipset related.

 

ps2cho: I think you're assuming a bit to much there. There's a difference between not knowing and just forgetting for the moment. I just wanted to remind you.

 

bump what program is everyone using to overclock the t5250

 

How long does it take usually from start to finish to do a pin mod?

 

You cant do it. No overclocking. He stopped posting cuz he couldnt do it.

 

oh ok thanks
i just wanted to poush it to 1.6

 

I have some more ideas that might merit discussion. I have an ASUS laptop with the same T5250 chip, and I want to overclock it. I've read ps2cho's posts (on other forums too), and I see that the pinmod didn't work as expected.

The Software FSB programs used to overclock make use of a 2 wire serial interface (SMBus) to control the PLL directly. This serial interface allows the PLL's internal registers to be modified. Each one is generally a byte, and the function of each register is model-specific. Each motherboard manufacturer can choose their own PLL, which usually generates many different clock signals (CPU, PCI, USB). The best way to find your PLL model is just to look at your motherboard, find the crystal (generally looks like a square or oval shaped shiny-top chip with about 4 contacts with a frequency listed on it - 14.318 in my case), and the PLL chip can't be too far away. I think the chip is usually smaller than a DRAM chip. Look up the model number on the internet, and you'll likely be able to find a data-sheet for it. Using one of the soft-FSB programs available you can set the registers manually, but generally they have presets for many different PLLs.

My PLL, the Silego SLG8SP510 doesn't have any registers that control the CPU frequency, so I figured I was simply out of luck. But the pinout and partial register layout matched another Silego PLL that was built into the soft-FSB program I was using, so I hoped that perhaps the registers were simply not documented for this part, and I could write them to overclock anyway. It turns out that is not the case, and writing the reserved registers did nothing.

So I thought I could replace the PLL with a pin-compatible model that did have the correct registers. I found a Silego model that had a pinout that was pretty close, but I didn't have the tools to solder such a small chip, and I don't know where I would find somebody to sell me one particular PLL chip.

So back to the hardware mods - the pinmod didn't work directly on the processor, but would it work on the PLL? The PLL has 3 frequency select pins, which act as inputs when the computer starts to select the bus frequency. The inputs are latched, then the pins can act as outputs if they are so configured. These pins directly correlate with the bsel[2:0] pins from the processor (the pins used in the pinmod).

So here's a theory, and tell me what you think: Intel or the BIOS developers are so stubborn that they take the multiplier and the bsel[2:0] pins information and compare it to the maximum qualified frequency of the CPU (read with CPUID). If the frequency is over the max qualified frequency, then the BIOS/hardware won't set the multiplier that high.

Now remember that the multiplier is controlled by Intel Speedstep and some MSRs have certain control over the multiplier. I use CrystalCPUID to edit the MSRs. There is a maximum bus ratio (multiplier) supported by a processor that can be read in MSR0x198[44:40]. The current operating ratio is given by MSR0x198[12:8]. Now that seems to be stuck at 6, but we want it at 9, right? For those who have tried the pinmod already, I suggest reading MSR0x199[12:8] - this value should be the current multiplier target. If you set it to 6, for example, then it will stay there forever. If you set it to 9, then the speedstep will use that as a multiplier target, so if the processor is not being used it may drop back to 6 (and this can be seen in MSR0x198), but under load it should go to 9. The lowest byte of MSR0x199 (bits [7:0]) represents voltage control, and I wouldn't mess with it unless you know what you're doing. Just use the value that was there when you read it.

So if my theory is right, then perhaps we can trick the logic into thinking that it's running slower than it really is. If the PLL simply generates a higher frequency (such as is done with the software FSB modifiers), then the CPU associated hardware/software will have no reason not to set the muliplier to 9 because it thinks it's running at 166 bus speed (based on the bsel[2:0] pins). I don't know if it can derive the ACTUAL frequency or not, so maybe this won't work?

But if I simply disconnect one of the frequency select pins on the PLL chip and drive it low or high (whichever will give me 200MHz), then it may push the CPU speed up without messing the multiplier. The only problem with my chip is, if I disconnect the pin, then my USB clock is screwed, because it's got both an input and output function.

So please respond with your comments.

 

Hi to everyone, 1st post here. anyways, i found something that may interest centrino duo oveclockers, i too have a t5250 on an hp pavilion dv 9640 clocked at 1.5ghz on a quanta 30cb mobo. i found a cpu-z validated log of someone overclocking his t5250 to 2160 Mhz . here's a link :

http://valid.x86-secret.com/show_oc.php?id=278602

it seems that he was using a samsung mobo and was able to oc the bus speed to 240 Mhz and rated fsb to 960 Mhz.

I am also planning to produce some more juice out of my processor by oc since the video card on this laptop is an 8600m gs and is already reaching 85C in game on stock speed of 500/1000/400 but cp is under 60C. any comments would be appreciated. thnx.

 

Have you noticed that hes running a ATI chipset ? Overclocking with those has benn easy compared to Intels chipset (no overclock at all)

 

Yep. For ATI chipsets -- this overclock works.

Intel chipsets -- No go!!

Revival!!

 

cannot give up. I'm shooting in the dark for my laptop since I haven't opened it up yet, but with setfsb and the plls it supports, I've been able to getfsb with three plls--- the slg* ones. that said, the first two immediately lock up or bsod. The third one causes programs to drop until it reboots, but I can usually get in a few setfsb's with that before it reboots on me. That said, each time it "sets" it doesn't actually set it.

My guess from not opening up the laptop is that the pll is indeed a slg* variant, but not one of the three.

Perhaps it's time to email the man behind setfsb and see if we can work something out.

Now I know this won't be of use to people that don't have my pll, but for me?

afaik, people have been setfsb'ing clevos, eeepcs, and others, and report that it indeed gets around the 6x lockdown limitation. Plus, we're finally at a point where pcie/pci can be adjusted independently, so it's reasonable to expect more than just a 3-5% boost. I won't be happy until I get to 200fsb at the least. Hoping for more than that though

 

I had a T5250 and this is what I did ...

bought a 7250 on ebay

 

If your cpu is an ES edition and has an unlocked multiplier could this work on an intel chipset?